Network-based Data Analysis And Its Applications On Metabolomics Study

In recent years,the development of system biology and complex network theory have provided a theoretical basis for the research of metabolic networks.By constructing the metabolic network,we could extract the important features in the network,which could provide new ideas for the research on metabolomics.In our study,we constructed an association network and applied it to tea data to get the metalmetabolite pairs which were strongly correlated.And then we proposed the way to construct a perturbation network based on the individual and applied it to colorectal cancer to select differential pathways.The main works includes:1.Association network modeling and its application in tea data.Firstly,we constructed the association networks by using Pearson correlation coefficients,regression model and Gaussian graphical model respectively with metals and metabolites in tea data.Secondly,we selected the metal-metabolite pairs with strong correlation according to the association network.Finally,the results of the three association networks were compared and we found that 8 combinations of metals and metabolites which were connected closely were selected by all three methods.In addition,we found that the association network based on regression model and Gaussian graphical model were more similar,and the association network based on regression model and Pearson correlation coefficient were very different.The combination of metabolites and metals in tea data with strong correlations could provide some reference for related research on tea.2.Perturbation network modeling based on the individual and its application in colorectal cancer data.Firstly,we used two methods to construct the association network based on the individual.Secondly,we constructed the perturbation network based on the individual.And then the perturbation of the association between two nodes in the network was calculated.Subsequently,the perturbation metabolite matrix based on the individual was constructed.That was,the perturbation of the correlation between two metabolites was converted into perturbation of corresponding metabolites by using the thresholds.Then,we constructed the perturbation pathway matrix based on the correspondence between metabolites and pathway.Finally,we used the statistical test method to select the pathways which were different significantly.Compared with the results of traditional methods,it showed that the differential pathways selected by the method proposed by this study could be used as a supplement to the traditional method.And the work we have done in this study may provide a new idea for the research of colorectal cancer.